the universal banking feedback effect: u.s. and canada evidence · 2015. 8. 19. · 1 cahier de...
TRANSCRIPT
1
Cahier de recherche 2015-06
The universal banking feedback effect: U.S. and Canada evidence
Christian Calmès Département des sciences administratives, Université du Québec (Outaouais), Campus St. Jérôme, 5 rue St Joseph, St Jérôme, Québec, Canada, J7Z 0B7; Chaire d’information
financière et organisationnelle (ESG UQAM).
Raymond Théoret Université du Québec (Montréal), École des sciences de la gestion, 315 est Ste-Catherine, R-3555, Montréal, Québec; Université du Québec à Montréal; Chaire d’information financière et organisationnelle, ESG UQAM.
�
(Fourth draft August 14th 2015)
2
The universal banking feedback effect: U.S. and Canada evidence
Abstract
Bank non-traditional business lines have significant feedback effects on both economic activity and the stock market. These effects appear in the non-interest income series after the 1997 structural break. Based on U.S. and Canadian data, we also find that the banking cycle tends to lead the business cycle in both countries over the whole sample period (first quarter of 1984 to last quarter of 2013), and particularly in the last decades. Despite these results, the impact of monetary policy on net interest income growth remains surprisingly stable after the emergence of universal banking. Keywords: Universal banking; Banking cycle; VAR; feedback effects. JEL classification: C32; G20; G21.
L’effet de rétroaction du banking universel :
.expérience américaine et canadienne
Les activités bancaires non traditionnelles exercent des effets de rétroaction importants sur l’activité économique et les marchés boursiers. Ces effets se manifestent dans les séries sur les revenus autres que d’intérêt à la suite du changement structurel de 1997. Sur la base des données américaines et canadiennes, nous trouvons aussi que le cycle bancaire tend à devancer le cycle économique dans les deux pays au cours de la période échantillonnale (premier trimestre de 1984 au dernier trimestre de 2013), et plus particulièrement durant les dernières décennies. En dépit de ces résultats, l’impact de la politique monétaire sur la croissance du revenu net d’intérêt demeure très stable à la suite de l’émergence du banking universel. Mots-clefs : Banking universel; cycle bancaire; VAR; effets de rétroaction. Classification JEL: C32; G20; G21.
3
1. Introduction
Since the revision of the Bank Act of 1987 and its subsequent amendments, the traditional four
pillars1 of Canadian banking have progressively been blurred by the rise of universal banking. In the U.S.,
commercial banks and investment bankers coexisted as separate entities until the subprime crisis but
commercial banks have been absorbing the investment banking industry since that time (Saunders et al.,
2014). The advent of universal banking may have important repercussions, both in terms of banking
cyclicality and the business cycle. For example, since the credit cycles tend to destabilize the economy
(e.g., Kiyotaki and Moore, 1997), universal banking could help mitigate this impact thanks to the
diversification benefits it entails (Calmès and Théoret, 2015). Indeed, fee-based activities might be less
cyclical than traditional banking and provide valuable buffer against fluctuations. On the other hand,
banking is also influenced by macroeconomic factors—such as GDP growth—and financial factors—like
interest rates and the stock market, and it is also possible that the impact of these factors on the various
components of banking activities—e.g., traditional versus fee-based—has changed.
The RBC models embedding financial intermediation feature a banking cycle related to firm’s
bankruptcy costs, and the impact of banks on the real sector is propagated through a credit supply effect
or a financial accelerator (Williamson 1987; Bernanke et al., 1999). More recent papers try to take also
into account bank non-traditional activities and assume that financial institutions are confronted with
financial or capital constraints aside non-financial borrowers (e.g., Gertler and Kiyotaki, 2011; Dewatcher
and Wouters, 2014). Interestingly, in this kind of models, the feedback effect from the banking to the real
sector may be substantial. In the same vein, some financial models put the emphasis on the tail-risk
(systemic risk) driven by diversification in off-balance-sheet activities such as securitization (e.g., Shleifer
and Vishny, 2010; Gennaioli et al., 2013).
However, on the empirical front, researchers tend to focus on the relationship between loan
defaults and the credit cycle, but they rarely incorporate fee-based activities explicitly (e.g., Bikker and
Hue, 2002; Jacobson et al., 2005; Marcucci and Quagliariello, 2006). In general, studies on bank fee-based
activities are more focused on the instability generated by these activities and the lack of overall
diversification they entail2, than on the relative cycles of traditional and fee-based activities per se (Stiroh,
2004; Stiroh and Rumble, 2006; Calmès et Liu, 2009; Calmès and Théoret, 2010, 2014). Following the
SVAR approach first introduced by Peersman and Wagner (2014), this paper aims at studying both the
U.S. and Canadian banking income cyclicality—and particularly the feedback effect these cycles may have
on the real economy—on a period stretching from 1984 to 2013. Peersman and Wagner (2014) find that
securitization indeed has an important feedback effect on U.S. real GDP over the period 1970-20083.
1 i.e., banking, insurance, fiduciary and brokerage activities. 2 Brown (2010) refers to the myth of diversification to qualify this situation. In fact, “overdivesification” in OBS activities reduces idiosyncratic risk but increases systematic risk which progresses towards systemic risk because of the interconnectedness between financial institutions overdiversification entails. Neglect of this systemic risk, or tail risk, evolves toward a financial crisis (Gennaioli et al., 2013). 3 According to Peersman and Wagner (2014), a securitization shock leads to a permanent rise of real GDP. Its effects are similar to a conventional technological shock in standard RBC models. The authors also find that an expansionary monetary policy shock leads to an increase in securitization in their model.
4
In this study instead, we concentrate on the three sources of banking cyclicality (shocks to GDP
growth, to the stock market and to the short-term interest rate) and our VAR experiments aim primarily
at gauging the feedback effect of banks’ fee-based activities on the real sector. Our main contribution is to
show that the feedback effects from banks’ non-traditional activities to the stock market are particularly
pronounced, and actually stronger than the ones associated with the real economy. We also find that the
feedback effect from non-traditional activities, to both the real sector and to the stock market, is more
significantly at play after 1997—this year corresponding to a structural break in the Canadian and U.S.
shares of non-interest income in net operating income (Calmès and Théoret, 2010)4. In line with the
standard bank duration gap model, our analysis supports the idea that net interest income reacts
negatively to the short-term interest rate. In other words, our results suggest that monetary policy
continues to impact net interest income in the same way, despite the multiplication of hedging vehicles
and the influence of universal banking.
This article is organized as follows. Section 2 presents the VAR methodology we use in this
paper. Section 3 provides the data sources and exposes the stylized facts related to the banking cycle in
both countries. In section 4 we analyze our VAR experiments before concluding in Section 5.
2. Methodology
We rely on VAR (vector autoregression) to study the impact of business cycles and financial
fluctuations on bank income flows. We compute the impulse response functions (IRF) of the two income
flows—i.e., net interest and non-interest—to shocks to the short-term interest rate, GDP growth and the
return of the stock market portfolio. An impulse response function is computed using a vector
autoregressive system (VAR) defined as follows: 1
n
i== + +∑t 0 i t -i tY A A Y ε
where Yt is the vector of
endogenous variables included in the VAR.
If the matrix of the residuals ( tε ) is diagonal, the
identification of the shocks associated with the variables is straightforward. For instance, if there are two
variables—say a real variable and financial variable—the residuals matrix provides directly on its
diagonal the respective shocks related to these variables—i.e., the real shock and the financial shock.
However, the residuals are usually correlated so that the residuals matrix is not diagonal. Hence, we need
a method to identify the shocks related to the variables of the VAR. Sims (1980) proposes the Cholesky
decomposition of the reduced form residuals’ covariance matrix, but to avoid the variables ordering
dependency, we rely on the generalized impulse response analysis (Pesaran and Shin, 1998). To build the IRF, we first transform the VAR equation into its infinite moving average
representation—i.e., an MA(∞ ) (Wold, 1938; Hamilton, 1994, p. 318-319). Then, we compute the partial
derivative of the MA(∞ ) as follows: 't
st
Yε
∂= Ψ
∂, where tε is the vector of innovations of the MA(∞ )
4 Peersman and Wagner (2014) do not investigate structural breaks in their sample.
5
representation. The last step consists in plotting the row i, column j element of sΨ —i.e., the ,i t s
jt
Yε+∂
∂.
This plot is the IRF. Each VAR is composed of the two income flows expressed in logarithmic differences
or growth rates, and of one of the three variables associated with shocks: the change in the short-term
interest rate5, the rate of growth of GDP and the return on the stock market portfolio. We compute these
VAR for the aggregate of U.S. and Canadian banks.
3. Data and stylized facts 3.1. Data
We estimate our VAR on two samples: the U.S. sample of all commercial banks, and the sample
of all Canadian banks. The U.S. banks’ statistics span the period ranging from the first quarter of 1984 to
the fourth quarter of 2013. These statistics are provided by the Federal Deposit Insurance Corporation
(FDIC). U.S. macroeconomic and financial time series are drawn from FRED, a database managed by the
Federal Reserve Bank of St-Louis. The Canadian banks’ sample comes from the Canadian Bankers
Association, the Office of Superintendent of Financial Institutions, and the Bank of Canada. Finally,
Canadian data on macroeconomic and financial time series are drawn from Cansim (from Statistics
Canada).
5 Since the series related to the short-term interest rate is not stationary, we include it in first differences in the VAR.
6
Figure 1 Net interest and non-interest income cycles: U.S. and Canada
Panel A: U.S.
Net interest Non-interest
Panel B: Canada
Net interest Non-interest
Notes: Shaded areas are associated with periods of economic slowdown. To compute the quarterly output gap, we first take the log of real GDP. We then detrend this transformed series with the Hodrick-Prescott filter using a smoothing coefficient (λ) equal to 1600—the trend of the series being a measure of potential output. The resulting residuals are the output gap measure.
3.2. Stylized facts
3.2.1 Banking cyclicality
Figure 1 provides the cycles of net interest and non-interest income growth computed with the
Hodrick-Prescott filter. U.S. banks show no obvious cycle for net interest income growth while, for
Canadian banks, the growth tends to be higher during recessions. Net interest income growth thus tends
-.12
-.08
-.04
.00
.04
.08
-.06
-.04
-.02
.00
.02
.04
1980 1985 1990 1995 2000 2005 2010
net interest income cycleoutput gap
-.4
-.3
-.2
-.1
.0
.1
.2
-.06
-.04
-.02
.00
.02
.04
1980 1985 1990 1995 2000 2005 2010
non-interest income cycleoutput gap
-.2
-.1
.0
.1
.2
-.06
-.04
-.02
.00
.02
.04
.06
1980 1985 1990 1995 2000 2005 2010
net interest income cycleoutput gap
-.6
-.4
-.2
.0
.2
.4
-.06
-.04
-.02
.00
.02
.04
.06
1980 1985 1990 1995 2000 2005 2010
non-interest income cycleoutput gap
7
to be countercyclical in Canada6. Regarding non-interest income growth, there is also more procyclicality
in the Canadian than in the U.S. series. These observations are supported by the spectral analysis of the
growth of bank income flows (Figure 2). In the U.S., non-interest income growth shows a moderate peak
at business cycle frequency—between 6 and 40 quarters (DeJong and Dave, 2007)—and another one at
the quarterly frequency. The spectrum of U.S. bank net interest income growth displays no obvious cycle
at business cycle frequency. It rather indicates that the series is persistent.
Figure 2 Spectral analysis of bank income components: U.S. and Canadian banks
U.S.
Canada
Notes: These spectra are built using an AR(p) model. Specifically, the spectrum is a decomposition of the variance of a time series by frequency—the cycle frequency being low near the origin and increasing progressively till π . Comparisons of the height of the spectrum for alternative values of frequency indicate the relative importance of fluctuations at the chosen frequencies in influencing variance of the time series. A spectrum having a peak near the origin indicates that the series is very persistent through time: its autocorrelation function declines very slowly. A spectrum having a peak in the shaded area indicates that this time series has a cycle in the conventional business cycle frequency. A spectrum which shifts to the left through time indicates less volatility for the series, i.e. a more stable series. Shaded areas correspond to the business cycle frequency, which is comprised between 6 and 40 quarters (DeJong and Dave, 2007).
6 Net interest income thus acts as a buffer during bad times.
0
10
20
30
40
50
60
70
0.0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2
frequency
spec
trum
U.S. banks' non-interest income growth spectrum
business cycle frequency
40 quarters 6 quarters
quarterly cycle
0
2
4
6
8
10
0.0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2
frequency
spec
trum
U.S. banks' net interest income growth spectrum
6 quarters40 quarters
business cycle frequency0
5
10
15
20
25
0.0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2
frequency
spec
trum
U.S. banks' net operating income growth spectrum
40 quarters 6 quarters
business cycle frequency
0
40
80
120
160
200
0.0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2
frequency
spec
trum
6 quarters40 quarters
business cycle frequency
quarterly cycle
Canadian banks' non-interest income growth spectrum
0
10
20
30
40
50
0.0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2
frequency
spec
trum
business cycle frequency
6 quarters40 quarters
quarterly cycle
Canadian banks's net interest income growth spectrum
0
10
20
30
40
50
0.0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2
frequency
spec
trum
business cycle frequency
40 quarters 6 quarters
quarterly cycle
Canadian banks' net operating income growth spectrum
8
In Canada, non-interest income growth has a clearer peak at business cycle frequency, which is
stronger than the U.S. one according to the ordinate of the spectrum. We also observe a peak at the
quarterly frequency. In contrast to the U.S., the spectrum of net interest income growth indicates that the
series has a standard cycle, but less pronounced than the one of non-interest income growth.
Table 1 Granger causality tests: U.S. and Canada
Notes: * indicates that the test is significant at the 10% level and ** indicates that the test is significant at the 5% level.
Test U.S. Canada
Real GDP growth GC non-interest income growth
F-Statistic 2.64 0.67
p-value 0.04** 0.61
Non-interest income growth GC real GDP growth
F-Statistic 3.98 3.47
p-value 0.01** 0.01**
Real GDP growth GC net interest income growth
F-Statistic 1.16 1.01
p-value 0.33 0.40
Net interest income growth GC real GDP growth
F-Statistic 0.93 0.96
p-value 0.44 0.43
Stock market return GC non-interest income growth
F-Statistic 0.98 3.32
p-value 0.41 0.01**
Non-interest income growth GC stock market return
F-Statistic 1.21 2.50
p-value 0.30 0.04**
Stock market return GC net interest income growth
F-Statistic 0.61 3.17
p-value 0.65 0.02**
Net interest income growth GC stock market return
F-Statistic 0.71 1.45
p-value 0.58 0.22
Change in T-bills rate GC non-interest income growth
F-Statistic 1.04 0.54
p-value 0.38 0.70
Non-interest income growth GC change in T-bills rate
F-Statistic 0.82 1.33
p-value 0.50 0.26
Change in T-bills rate GC net interest income growth
F-Statistic 3.77 2.36
p-value 0.01** 0.05**
Net interest income growth GC change in T-bills rate
F-Statistic 1.63 1.07
p-value 0.17 0.37
9
3.2.2 Granger causality tests
VAR embed many interactions between the selected variables, and this ought to be first captured
by Granger causality tests (Table 1). Our tests suggest that GDP growth Granger causes non-interest
income growth only for U.S. banks, but the reverse also holds true in both countries, the tests being
significant at the 1% level. Overall, the growth of non-interest income seems to lead GDP growth, a
prima facie evidence of a feedback effect from the banking sector to the real economy. Moreover, in
Canada, stock market return Granger causes non-interest income growth and the reverse, albeit less
significant, is also true. Granger causality tests related to the short-term interest rate are negative for
both countries, suggesting that non-interest income growth is not very sensitive to changes in the short-
term interest rate.
Net interest income growth does not lead any of the three selected shocks. In contrast to non-
interest income, the feedback effects from the traditional banking activities to the real sector do not seem
to be operative. In both countries, tests are negative when linking net interest income growth to GDP
growth. Concerning tests involving the stock market return, note that this variable Granger causes net
interest income growth only in Canada. Finally, as well documented in the literature, a change in the
short-term interest rate seems to Granger cause net interest income growth, the test being more
conclusive in the U.S. (an increase in the interest rate leads to a decrease in net interest income growth).
Figure 3 Cross-correlations: U.S. and Canada
Non-interest income growth and GDP growth
U.S. Canada
G4_NONIN,G4_REALGDP...G4_NONIN,G4_REALGDP... i lag lead
0 0.2262 0.22621 0.1186 0.30972 0.1424 0.42403 0.2007 0.48094 0.2349 0.49665 0.3016 0.45116 0.2810 0.34217 0.2501 0.24498 0.1998 0.15719 0.1610 0.1531
10 0.1064 0.122811 0.0726 0.133012 0.1068 0.066213 0.0822 -0.015614 0.0808 -0.013615 0.0454 -0.045216 0.0453 0.004917 0.0425 0.026718 0.0188 0.033319 0.0373 0.032420 -0.0020 0.002221 0.0220 -0.038322 0.0634 -0.094323 0.0516 -0.086624 0.0625 -0.063625 0.0148 -0.010126 -0.0130 0.072627 -0.0324 0.089028 -0.0367 0.110129 -0.0810 0.114330 -0.1329 0.088631 -0.1591 0.074132 -0.1809 0.040433 -0.1390 0.022734 -0.0899 0.002935 -0.0068 -0.018836 0.0007 -0.0733
G4_NOI,G4_PIBREAL(-i) G4_NOI,G4_PIBREAL(+i) i lag lead
0 0.1105 0.11051 0.0190 0.22792 -0.0023 0.32033 0.0079 0.42134 0.0593 0.43205 0.0787 0.36416 0.0600 0.27397 0.0209 0.16168 -0.0345 0.06829 -0.0362 0.0458
10 -0.0066 0.045411 0.0018 0.010012 0.0196 -0.004313 0.0265 -0.059014 0.0159 -0.119315 0.0137 -0.108016 0.0203 -0.081517 0.0013 -0.025218 -0.0161 0.038419 -0.0198 0.051520 -0.0402 0.028121 -0.0688 -0.003422 -0.0848 -0.034823 -0.0963 -0.015024 -0.1116 0.029625 -0.0862 0.079126 -0.0857 0.124927 -0.1314 0.114228 -0.1627 0.099929 -0.2268 0.086730 -0.2686 0.067231 -0.2638 0.063432 -0.2722 0.047333 -0.2400 0.012234 -0.2073 -0.023135 -0.1453 -0.055936 -0.0576 -0.0959
10
Net interest income growth and GDP growth
U.S. Canada
Net operating income growth and GDP growth
U.S. Canada
G4_NETIN,G4_REALGDP... G4_NETIN,G4_REALGDP... i lag lead
0 0.1445 0.14451 0.0721 0.13312 -0.0086 0.08103 -0.0871 0.01894 -0.1184 -0.03505 -0.0924 -0.00356 0.0090 0.04707 0.1134 0.11078 0.2108 0.15589 0.2803 0.1440
10 0.2975 0.116211 0.3027 0.090612 0.2865 0.079713 0.2986 0.076514 0.3162 0.081715 0.3445 0.069116 0.3404 0.020217 0.2837 -0.033618 0.2112 -0.109419 0.1300 -0.168320 0.0879 -0.187921 0.0489 -0.182822 0.0180 -0.142123 -0.0150 -0.098424 -0.0427 -0.062825 -0.0447 -0.054526 -0.0467 -0.067427 -0.0389 -0.083028 -0.0221 -0.078729 -0.0113 -0.058930 0.0038 -0.009231 -0.0059 0.032632 -0.0295 0.053933 -0.0564 0.075534 -0.0750 0.070635 -0.0666 0.058236 -0.0362 0.0246
G4_NI,G4_PIBREAL(-i) G4_NI,G4_PIBREAL(+i) i lag lead
0 -0.2809 -0.28091 -0.2258 -0.31112 -0.0895 -0.29443 0.0224 -0.21904 0.1009 -0.14935 0.1528 -0.07796 0.1338 -0.03907 0.1071 -0.03548 0.0657 -0.03049 0.0177 -0.0159
10 -0.0068 0.041711 -0.0345 0.092612 -0.0322 0.111313 0.0261 0.100714 0.0595 0.056315 0.1019 0.017816 0.1324 -0.015217 0.0779 -0.028618 0.0150 -0.053719 -0.0671 -0.072420 -0.1998 -0.033721 -0.2598 -0.008922 -0.2590 0.046923 -0.2439 0.114124 -0.1487 0.121125 -0.0628 0.125526 -0.0228 0.083427 0.0302 -0.014328 0.0639 -0.119429 0.0695 -0.203230 0.0871 -0.211431 0.0982 -0.143632 0.0818 -0.028133 0.0998 0.085334 0.1008 0.148135 0.0758 0.174336 0.0741 0.1637
G4_NETINCOME,G4_REA... G4_NETINCOME,G4_REA... i lag lead
0 0.2636 0.26361 0.1416 0.31382 0.1066 0.36223 0.0962 0.36544 0.0980 0.34565 0.1587 0.33986 0.2067 0.29137 0.2525 0.25548 0.2828 0.21409 0.3038 0.1919
10 0.2792 0.150111 0.2595 0.139412 0.2707 0.087513 0.2593 0.032514 0.2667 0.038515 0.2572 0.011116 0.2512 0.014417 0.2120 -0.005018 0.1527 -0.049419 0.1146 -0.088720 0.0634 -0.126221 0.0548 -0.155022 0.0601 -0.169823 0.0307 -0.138624 0.0207 -0.096425 -0.0107 -0.049526 -0.0313 0.001227 -0.0401 0.007128 -0.0382 0.029329 -0.0714 0.045930 -0.1019 0.061431 -0.1301 0.076232 -0.1593 0.062633 -0.1423 0.061634 -0.1154 0.043135 -0.0505 0.017736 -0.0232 -0.0465
G4_NETINC,G4_PIBREAL...G4_NETINC,G4_PIBREAL... i lag lead
0 -0.0335 -0.03351 -0.0878 0.05472 -0.0405 0.14673 0.0239 0.27424 0.1043 0.32455 0.1391 0.30216 0.1026 0.23837 0.0444 0.13588 -0.0338 0.04379 -0.0540 0.0238
10 -0.0305 0.043911 -0.0266 0.025812 -0.0008 0.020713 0.0336 -0.036114 0.0356 -0.107415 0.0520 -0.103716 0.0705 -0.086217 0.0304 -0.030618 -0.0051 0.022619 -0.0364 0.027120 -0.1071 0.022721 -0.1506 0.005722 -0.1613 0.003223 -0.1646 0.053224 -0.1322 0.100625 -0.0742 0.145826 -0.0603 0.166027 -0.0846 0.112428 -0.1077 0.048829 -0.1749 0.001030 -0.2155 -0.019431 -0.2205 0.003832 -0.2451 0.033433 -0.2111 0.041534 -0.1837 0.028535 -0.1313 0.002836 -0.0461 -0.0415
11
Non-interest income growth and stock market return
U.S. Canada
Net interest income growth and stock market return
U.S. Canada
G4_NONIN,RSP4_500(-i) G4_NONIN,RSP4_500(+i) i lag lead
0 0.2677 0.26771 0.0599 0.37302 -0.0129 0.40633 -0.0099 0.36494 0.0078 0.31135 0.1487 0.27056 0.2205 0.18107 0.2417 0.08828 0.2506 -0.01169 0.2686 -0.1134
10 0.2097 -0.211211 0.1649 -0.217912 0.1182 -0.224413 -0.0032 -0.218514 -0.0203 -0.098215 -0.0622 -0.024116 -0.0558 -0.027217 0.0036 -0.012018 0.0014 -0.041819 0.0343 -0.099420 0.0906 -0.142821 0.1220 -0.152822 0.1841 -0.147823 0.2347 -0.093024 0.1993 0.001625 0.1765 0.006926 0.1118 0.023527 0.0756 0.009628 0.0821 0.012029 0.0873 0.063230 0.0662 0.055131 0.0391 0.066832 0.0264 0.045933 -0.0112 0.009434 0.0716 -0.001535 0.0982 -0.067536 0.0892 -0.1146
G4_NOI,RTSX4(-i) G4_NOI,RTSX4(+i) i lag lead
0 0.3303 0.33031 0.1854 0.36112 -0.0409 0.42373 -0.0898 0.36744 -0.1516 0.24075 -0.1475 0.08686 -0.0297 -0.08107 -0.0467 -0.16658 -0.0733 -0.15309 -0.0110 -0.1257
10 -0.0407 -0.113211 -0.0160 -0.125612 0.0629 -0.169413 0.0492 -0.134914 0.0946 -0.039715 0.0971 0.027616 0.0222 0.083117 -0.0150 0.062018 -0.0488 -0.038819 -0.0582 -0.137520 -0.0037 -0.176621 0.0090 -0.181722 0.0533 -0.111823 0.0840 0.026924 0.0717 0.102325 0.0947 0.198026 0.0750 0.228427 0.0418 0.155728 -0.0409 0.131229 -0.1106 0.090930 -0.2407 0.068131 -0.3019 0.079232 -0.2744 0.050233 -0.2526 0.020634 -0.1403 -0.024535 -0.0402 -0.067036 0.0486 -0.0860
G4_NETIN,RSP4_500(-i) G4_NETIN,RSP4_500(+i) i lag lead
0 0.0235 0.02351 0.0188 0.01412 0.0005 -0.02223 -0.0534 -0.02774 -0.0857 -0.04555 -0.0764 -0.04176 -0.0322 -0.00997 0.0211 0.01638 0.0749 0.02819 0.1373 -0.0045
10 0.1609 -0.052611 0.1594 -0.100112 0.1277 -0.070213 0.1056 -0.022214 0.1240 0.016815 0.1587 0.064616 0.1943 0.041017 0.1824 0.020718 0.1323 0.028219 0.0780 0.036520 0.0472 0.055121 0.0306 0.090322 0.0341 0.098523 0.0477 0.089824 0.0453 0.076525 0.0449 0.030726 0.0286 -0.014327 -0.0171 -0.059228 -0.0645 -0.080229 -0.0975 -0.066330 -0.1177 -0.022331 -0.0838 0.037532 -0.0243 0.071133 0.0170 0.067534 0.0451 0.032235 0.0471 -0.002936 0.0491 -0.0372
G4_NI,RTSX4(-i) G4_NI,RTSX4(+i) i lag lead
0 -0.3936 -0.39361 -0.4369 -0.34932 -0.3624 -0.28593 -0.1730 -0.19674 -0.0081 -0.17385 0.1597 -0.10316 0.2801 -0.05197 0.2630 -0.02148 0.2772 0.06169 0.2485 0.0508
10 0.1532 0.067811 0.0683 0.060112 -0.0363 0.001513 -0.0936 0.027214 -0.0653 0.071115 0.0107 0.094716 0.0673 0.111417 0.1396 0.109518 0.2015 0.061019 0.1958 0.050220 0.1694 0.092121 0.0888 0.100322 -0.0515 0.127723 -0.1200 0.145324 -0.1540 0.084525 -0.1538 0.035026 -0.1150 -0.057127 -0.1392 -0.187928 -0.1312 -0.252929 -0.1432 -0.258330 -0.1269 -0.184231 -0.0245 -0.033832 0.0447 0.109533 0.1440 0.169734 0.2012 0.216735 0.1359 0.210236 0.0639 0.1503
12
Net operating income growth and stock market return
U.S. Canada
Non-interest income growth and change in the T-bill rate
U.S. Canada
G4_NETINCOME,RSP4_5...G4_NETINCOME,RSP4_5... i lag lead
0 0.1951 0.19511 0.0430 0.27202 -0.015... 0.28663 -0.042... 0.26424 -0.043... 0.21835 0.0708 0.18576 0.1508 0.12517 0.1967 0.06358 0.2322 -0.009...9 0.2798 -0.105...
1... 0.2544 -0.205...1... 0.2223 -0.244...1... 0.1712 -0.226...1... 0.0699 -0.189...1... 0.0625 -0.072...1... 0.0527 0.02001... 0.0759 0.00941... 0.1108 0.00941... 0.0842 -0.006...1... 0.0795 -0.042...2... 0.1075 -0.070...2... 0.1238 -0.059...2... 0.1644 -0.052...2... 0.2019 -0.020...2... 0.1712 0.04442... 0.1514 0.02572... 0.1012 0.01262... 0.0515 -0.021...2... 0.0249 -0.028...2... 0.0072 0.01723... -0.022... 0.03523... -0.021... 0.07593... 0.0080 0.07683... 0.0079 0.04273... 0.0818 0.01253... 0.1006 -0.054...3... 0.0970 -0.113...
G4_NETINC,RTSX4(-i) G4_NETINC,RTSX4(+i) i lag lead
0 0.1508 0.15081 -0.0163 0.20802 -0.2052 0.30383 -0.1705 0.29314 -0.1666 0.18025 -0.0812 0.04906 0.0859 -0.10227 0.0556 -0.18658 0.0454 -0.14939 0.0855 -0.1238
10 0.0186 -0.107911 0.0232 -0.124012 0.0484 -0.183313 0.0120 -0.139714 0.0636 -0.017715 0.0853 0.068716 0.0339 0.131717 0.0294 0.117118 0.0167 0.005519 0.0176 -0.087320 0.0742 -0.094121 0.0574 -0.090422 0.0582 -0.023123 0.0576 0.105924 0.0307 0.129925 0.0562 0.189726 0.0450 0.183727 -0.0044 0.065728 -0.0895 0.026529 -0.1679 -0.009730 -0.2869 -0.003831 -0.3021 0.058232 -0.2470 0.085933 -0.1844 0.077934 -0.0500 0.054035 0.0226 0.015236 0.0854 -0.0349
G4_NONIN,DTB_3MOIS(-i... G4_NONIN,DTB_3MOIS(+... i lag lead
0 0.0285 0.02851 -0.0017 0.10432 0.0263 0.12983 0.0408 0.11944 -0.0875 0.19965 -0.0285 0.10816 -0.0673 0.01787 -0.1748 0.00188 -0.1095 0.01829 -0.0020 -0.0065
10 0.0230 -0.051511 0.0108 -0.046712 0.0694 -0.010413 0.0984 -0.021914 0.1140 -0.019915 0.0968 -0.035516 0.0812 -0.086717 0.0016 -0.132218 -0.0383 -0.065519 0.0722 -0.089220 -0.0429 -0.131121 -0.0604 -0.075922 -0.0429 -0.126423 -0.0759 -0.089224 -0.0432 0.021725 -0.0706 0.060826 0.0020 0.111327 -0.0224 0.149828 -0.0451 0.174029 -0.0813 0.141930 -0.0832 0.079831 -0.0966 0.089632 -0.2255 0.064733 -0.0907 0.048734 -0.0689 0.040635 -0.0031 0.009436 0.2369 -0.0381
G4_NOI,DTB_3MOIS(-i) G4_NOI,DTB_3MOIS(+i) i lag lead
0 0.1496 0.14961 -0.0130 0.20352 -0.0131 0.21073 -0.1048 0.16564 -0.1569 0.14055 -0.0415 0.06826 -0.0729 0.01527 -0.1022 0.01788 -0.0372 -0.04279 -0.0945 -0.0037
10 -0.0799 0.011111 0.0212 0.003812 -0.0062 0.014613 0.0613 -0.076114 0.0541 -0.103915 -0.0664 -0.096416 0.0261 -0.096917 -0.0325 -0.114418 -0.0745 -0.086919 0.0385 -0.039820 -0.0893 -0.029921 0.0091 -0.000922 0.0680 -0.011523 -0.0247 0.019424 -0.0312 0.066225 -0.0205 0.093526 -0.0492 0.083527 0.0366 0.052828 0.0754 -0.009129 -0.0589 -0.002030 -0.0374 -0.017531 -0.1136 -0.052932 -0.1016 -0.046033 -0.0617 -0.027634 -0.1262 -0.009435 0.0139 0.019136 -0.0269 -0.0073
13
Net interest income growth and change in the T-bill rate
U.S. Canada
Net operating income growth and change in the T-bills rate
U.S. Canada
Notes: Each panel provides the correlations of the first factor with the lags and leads of the second factor. The confidence intervals of the correlations appear on the plots.
G4_NETIN,DTB_3MOIS(-i) G4_NETIN,DTB_3MOIS(+i... i lag lead
0 -0.1869 -0.18691 -0.2262 -0.19262 -0.2771 -0.21723 -0.3588 -0.16754 -0.2933 -0.09515 -0.1608 0.02836 -0.0221 0.07767 0.0967 0.10408 0.0971 0.13359 0.0822 0.1585
10 0.0418 0.218611 0.0672 0.249212 0.1355 0.255413 0.1488 0.208314 0.1519 0.125715 0.1427 0.057316 0.1912 0.035817 0.2197 -0.005118 0.2149 -0.029719 0.1701 -0.055020 0.0524 -0.101021 0.0114 -0.143922 0.0065 -0.165523 -0.0644 -0.175624 -0.1688 -0.248325 -0.2277 -0.251426 -0.2896 -0.216327 -0.2477 -0.200228 -0.1603 -0.057129 -0.0930 0.053930 -0.0052 0.109631 0.0120 0.154432 0.0202 0.137033 -0.0082 0.122534 -0.0284 0.127135 -0.0539 0.138036 -0.0716 0.1225
G4_NI,DTB_3MOIS(-i) G4_NI,DTB_3MOIS(+i) i lag lead
0 -0.2329 -0.23291 -0.2316 -0.01542 -0.2609 0.01943 -0.2993 -0.00414 -0.0671 0.14435 0.0378 0.10666 0.0743 0.01747 0.1645 0.05598 0.1046 -0.05159 0.1013 -0.0463
10 0.1242 0.074511 0.1002 0.037212 0.0869 0.084013 0.0578 0.140314 0.0112 0.077415 0.0370 0.125816 0.0645 0.092617 0.0877 -0.058418 0.1069 -0.035519 -0.0224 -0.081820 -0.0797 -0.119621 -0.0806 -0.016922 -0.1331 -0.039223 -0.0819 -0.016524 -0.0197 0.077925 -0.0403 0.029626 0.0257 0.021327 0.0528 -0.046628 0.0708 -0.143129 0.0669 -0.165330 0.0818 -0.164431 0.0524 -0.142832 0.0086 -0.054733 -0.0474 -0.006034 -0.0550 0.066135 0.0255 0.172936 -0.0203 0.1277
G4_NETINCOME,DTB_3M...G4_NETINCOME,DTB_3M... i lag lead
0 -0.0898 -0.08981 -0.1329 -0.04472 -0.1547 -0.05343 -0.2178 -0.03064 -0.2670 0.08005 -0.1354 0.10056 -0.0664 0.06317 -0.0589 0.06768 -0.0172 0.09769 0.0442 0.0916
10 0.0340 0.104311 0.0532 0.134512 0.1476 0.160513 0.1768 0.124114 0.1901 0.070815 0.1655 0.005216 0.1895 -0.042417 0.1508 -0.103418 0.1151 -0.077919 0.1634 -0.106520 -0.0006 -0.164121 -0.0416 -0.150422 -0.0279 -0.188323 -0.0881 -0.162924 -0.1380 -0.137125 -0.1964 -0.114626 -0.1792 -0.061027 -0.1781 -0.030028 -0.1420 0.087029 -0.1235 0.142530 -0.0797 0.133031 -0.0822 0.168932 -0.1557 0.135633 -0.0715 0.109134 -0.0621 0.106635 -0.0248 0.093536 0.1338 0.0499
G4_NETINC,DTB_3MOIS(... G4_NETINC,DTB_3MOIS(... i lag lead
0 0.0376 0.03761 -0.0874 0.18222 -0.1123 0.20583 -0.2174 0.14534 -0.1564 0.18845 -0.0242 0.10666 -0.0380 -0.00137 -0.0154 0.02408 -0.0036 -0.07099 -0.0399 -0.0387
10 -0.0112 0.038811 0.0599 0.006612 0.0327 0.037113 0.0715 -0.004114 0.0325 -0.059415 -0.0626 -0.029716 0.0377 -0.051417 -0.0059 -0.142218 -0.0170 -0.103519 0.0236 -0.078220 -0.1123 -0.084921 -0.0083 -0.003722 0.0100 -0.011223 -0.0443 0.029124 -0.0115 0.127025 -0.0292 0.127826 -0.0245 0.118527 0.0619 0.050828 0.0906 -0.052529 -0.0384 -0.069730 -0.0080 -0.085031 -0.0993 -0.109532 -0.1060 -0.066333 -0.0964 -0.034234 -0.1600 0.013435 0.0099 0.097636 -0.0546 0.0552
14
3.2.3 Cross-correlations between bank income flows and macroeconomic and financial variables
Cross-correlations between bank income flows variables, on the one hand, and the set of
macroeconomic and financial variables, on the other hand, are an important input in the preliminary
interpretation of our VARs. In the U.S., non-interest income growth is more positively correlated with
leads of GDP growth than with its lags, although the correlation with lags is significant and quite
persistent (Figure 3). In Canada, the cross-correlation between non-interest income growth and lags in
GDP growth is not observed in the short-run, and the positive correlation with leads of GDP growth is
quite high but less persistent. These properties are further evidence that there might indeed exist a
feedback effect from bank non-interest income growth to the real sector in both countries.
Turning to the stock market return, note that the positive correlation of U.S. non-interest income
growth with lags of this variable is significant during twelve quarters, but the positive correlation with
leads is much higher (albeit shorter). Not surprisingly, the waves are less persistent than for GDP
growth. In Canada, non-interest income growth is correlated positively with the stock market return and
its one-quarter lag. The profile is thus less persistent than in the U.S., and it reverses more quickly. The
positive correlation between non-interest income growth and the stock market return leads is higher than
in the U.S. (it is also less persistent than in the case of GDP growth). These results corroborate our
previous findings, as the cross-correlation profiles suggest the presence of a feedback effect from non-
interest income growth to the stock market in both countries.
Finally, note that the cross-correlation between non-interest income growth and the short-term
interest rate is much lower than with GDP growth or the stock market return. More precisely, there is a
positive feedback effect from non-interest income growth to the short-term interest rate—especially in
Canada—but this feedback effect is much weaker than in the case of GDP growth or the stock market. In
other respects however, the correlation between U.S. net interest income growth and the lagged short-
term rate has a profile which is consistent with the standard bank duration gap model7, being negative for
the first lagged quarters and positive thereafter. In Canada, there is also a negative correlation between
net interest income growth and lagged interest rate which lasts over four quarters. The correlation is
lower in Canada and we observe no obvious correlation with the leads of the interest rate in Canada.
Net interest income growth is correlated positively and significantly with GDP growth,
beginning at the t-6 lag in the U.S., and this correlation is still persistent for remote lags. In contrast, in
Canada, the correlation of net interest income growth is negative over periods t and t-1. We observe also
a significant negative correlation for leads t+1 to t+4. This counterintuitive profile is consistent with the
negative correlation between the two components of net operating income in Canada, which reflects a
different bank product-mix. For the same reason, the correlation profile between U.S. net interest income
growth and the stock market return is similar to GDP growth, whereas in Canada net interest income
7 Since banks have usually a positive duration gap, the duration of assets being greater than the durations of liabilities.
15
growth is correlated negatively with the stock market return over four-quarter lags and leads. Not
surprisingly, the correlation is also higher than with GDP growth. Overall, the relationship between net
interest income growth and GDP growth is more in line with the traditional banking channel in the U.S.
than in Canada—an increase in real GDP growth being associated with a rise in lending. More
importantly, the feedback effects from net interest income growth to real GDP growth or the stock
market are weak or non-existent.
Summarizing, we arrive at a two-part phenomenon: (i) net interest income is influenced by the
macroeconomic shocks, but do not lead them; (ii) Non-interest impacts GDP and especially the stock
market and is not much influenced by the interest rate. These are the two stylized facts motivating the
VAR analysis presented in the next section.
4. VAR analysis
An IRF plots the impact of a one-time shock to one of the innovations of the variables which
constitute the VAR system. A shock is equal to one standard deviation of the innovation. The variables
which constitute our VAR analysis are: bank non-interest income growth; bank net interest income
growth; bank net operating income growth—net operating income being the sum of net interest income
and non-interest income; real GDP growth; the stock market return—i.e., the S&P500’s in the U.S. and
the S&P/TSX in Canada; the short-term interest rate—i.e., the three-month Treasury bills rate.
Figure 4 IRFs: Growth of income components and GDP growth
U.S.
Canada
-4
-2
0
2
4
6
8
10
2 4 6 8 10 12 14 16 18 20
Response of non-interest income growth to GDP growth
-2
-1
0
1
2
3
2 4 6 8 10 12 14 16 18 20
Response of net interest income growth to GDP growth
-0.8
-0.4
0.0
0.4
0.8
1.2
2 4 6 8 10 12 14 16 18 20
Response of real GDP growth to non-interest income growth
-10
-5
0
5
10
15
2 4 6 8 10 12 14 16 18 20
Response of non-interest income growth to GDP growth
-4
-2
0
2
4
6
2 4 6 8 10 12 14 16 18 20
Response of net interest income growth to GDP growth
-0.50
-0.25
0.00
0.25
0.50
0.75
1.00
1.25
1.50
2 4 6 8 10 12 14 16 18 20
response of GDP growth to non-interest income growth
16
4.1. Interactions between the growth of income flows and real GDP growth
Figure 4 provides the IRFs linking the two components of net operating growth and real GDP
growth in the U.S. and in Canada. In the U.S., non-interest income growth reacts positively and
significantly to a GDP growth shock, but this impact is short-lived while in Canada there is no significant
impact. Regarding net interest income growth, there is a lagged significant non-linear response to the
shock in the U.S. In Canada, the corresponding response is again not significant. Therefore, banks’
income flows seem slightly sensitive to the lagged values of real GDP growth.
More importantly, the response of real GDP growth to a non-interest income growth shock is
positive and significant in both countries8. This effect is quite persistent since it lasts eight quarters.
Consistent with the Granger causality tests, this result confirms the presence of a significant feedback
effect from non-interest income growth to GDP growth, and this effect is stronger than the traditional
direct effect (from the real sector to the banking industry).
Figure 5 IRFs: Growth of income components and GDP growth (4-quarter lead)
U.S.
Canada
As a robustness check, Figure 5 provides the IRFs associated with a lead of four quarters to GDP
growth9. In both countries, non-interest income growth still reacts positively and significantly to the lead
8 Note that all variables of our VAR system are endogenous—i.e., real GDP growth, the stock market return and the short-term interest rate are endogenous aside the components of net operating income growth. 9 More precisely, we regress the growth of banks’ income flows on a lead of four quarters in GDP growth. This lead follows from the analysis of the cross-correlation plots.
-4
-2
0
2
4
6
8
2 4 6 8 10 12 14 16 18 20
Response of non-interest income growth to real GDP growth (lead 4 quarters)
-2
-1
0
1
2
3
2 4 6 8 10 12 14 16 18 20
Response of net interest income growth to real GDP growth (lead 4 quarters)
-10
-5
0
5
10
15
2 4 6 8 10 12 14 16 18 20
Response of non-interest income growth to real GDP growth (lead 4 quarters)
-4
-2
0
2
4
6
2 4 6 8 10 12 14 16 18 20
Response of net interest income growth to real GDP growth (lead 4 quarters)
17
of GDP growth, this impact being stronger in Canada than in the U.S. By contrast, net interest income
growth does not respond significantly to this lead. In other words, if there is a significant impact of
banking fluctuations on the real economy, it is more likely stemming from off-balance-sheet activities
than from the loan business.
Figure 6 IFRs: Growth of income components and stock market return
U.S.
Canada
-4
-2
0
2
4
6
8
10
12
2 4 6 8 10 12 14 16 18 20
Response of non-interest income growth to S&P500 return
-2
-1
0
1
2
3
2 4 6 8 10 12 14 16 18 20
Response of net interest income growth to S&P500 return
-8
-4
0
4
8
12
2 4 6 8 10 12 14 16 18 20
Response of S&P500 return to non-interest income growth
-10
-5
0
5
10
15
2 4 6 8 10 12 14 16 18 20
Response of non-interest income growth to TSX return
-4
-2
0
2
4
6
2 4 6 8 10 12 14 16 18 20
Response of net interest income growth to TSX return
-8
-4
0
4
8
12
2 4 6 8 10 12 14 16 18 20
Response of TSX return to non-interest income growth
18
4.2. Interactions between the growth of income flows and the stock market return
In the U.S., non-interest income growth also reacts positively and significantly to a shock to the
stock market return, but this effect is shorter-lived (Figure 6). It is stronger in Canada. On the other
hand, net interest income growth shows no significant response to the same shock in both countries.
More importantly, and in line with our previous finding, non-interest income growth impacts positively
and significantly the stock market returns. In Canada, this feedback effect is actually stronger than the
one associated with GDP growth, albeit shorter-lived. Hence, there seems to exist a feedback effect from
banks’ fee-based activities to the stock market.
Figure 7 IRFs: Growth of income components and stock market return (4-quarter lead)
U.S.
Canada
-4
-2
0
2
4
6
8
10
12
2 4 6 8 10 12 14 16 18 20
Response of non-interest income growth to S&P500 return (lead 4 quarters)
-2
-1
0
1
2
3
2 4 6 8 10 12 14 16 18 20
Response of net interest income growth to S&P500 return (lead 4 quarters)
-10
-5
0
5
10
15
2 4 6 8 10 12 14 16 18 20
Response of non-interest income growth to TSX return (lead 4 quarters)
-4
-3
-2
-1
0
1
2
3
4
2 4 6 8 10 12 14 16 18 20
Response of net interest income growth to TSX return (lead 4 quarters)
19
The IRFs with a lead of four quarters to the stock market return corroborate this result.
Consistent with the response of the stock market return to non-interest income growth, the response of
non-interest income growth to the lead of the stock market return is stronger and more significant in
Canada than in the U.S. (Figure 7). This is consistent with the different product-mix of the two banking
systems, Canadian banks focusing more on market-based activities than their U.S. counterparts (Calmès
and Théoret, 2015).
Overall, this set of results again supports the idea that there might well be a feedback effect, with
banking fluctuations significantly leading stock market fluctuations.
Figure 8 IRFs: Growth of income components and change in the T-bills rate
U.S.
Canada
4.3. Interactions between the growth of income flows and the short-term interest rate
Consistent with the plots of the cross-correlations, the response of non-interest income growth
to a shock to the short-term interest rate is not significant in both banking systems (Figure 8). However,
-4
-2
0
2
4
6
8
10
2 4 6 8 10 12 14 16 18 20
Response of non-interest income growth to 3month Treasury bills
-2
-1
0
1
2
3
2 4 6 8 10 12 14 16 18 20
Response of net interest income growth to 3month Treasury bills
-10
-5
0
5
10
15
2 4 6 8 10 12 14 16 18 20
Response of non-interest income growth to 3month Treasury bills
-4
-2
0
2
4
6
2 4 6 8 10 12 14 16 18 20
Response of net interest income growth to 3month Treasury bills
20
following a positive shock to the short-term interest rate, net interest income growth decreases in both
countries, as expected. This result is consistent with the standard duration gap model, as the duration of
banks’ assets is longer than the duration of their liabilities—i.e., banks fund long-term assets with
deposits and liabilities which have a shorter duration than assets. According to the IRFs’ profiles,
monetary policy seems to impact more the income associated with traditional activities in the U.S. than in
Canada (which could partly be attributable to the price stability targeting versus inflation targeting
policies of the two countries).
Figure 9 IRFs: Net operating income growth and various shocks
U.S.
GDP growth (no lead) GDP growth (4-quarter lead)
Stock market return (no lead) Stock market return (4-quarter lead)
Change in the T-bills rate
-2
-1
0
1
2
3
4
2 4 6 8 10 12 14 16 18 20
Response of net income growth to real GDP growth (no lead)
-1
0
1
2
3
4
2 4 6 8 10 12 14 16 18 20
Response of net income growth to GDP growth (4-quarter lead)
-2
-1
0
1
2
3
4
2 4 6 8 10 12 14 16 18 20
Response of net income growth to the stock market return (no lead)
-2
-1
0
1
2
3
4
2 4 6 8 10 12 14 16 18 20
Response of net income growth to the stock market return (4-quarter lead)
-2
-1
0
1
2
3
4
2 4 6 8 10 12 14 16 18 20
Response of net income growth to the change in the T-bills rate
21
Finally, Figure 9 displays the response of net operating income growth to various shocks for the
two banking systems. It is interesting to note that in both countries, net operating income growth reacts
positively and significantly to the leaded values of GDP growth and the stock market return. The
responses of net operating income growth to the lagged values of GDP growth are lower and less
persistent in both countries. Regarding the interest rate shock, we observe a decrease in net operating
income growth in the U.S. while no such effect is observed in Canada.
Figure 10 Quandt-Andrews unknown breakpoint test on U.S. banks’ snonin.
U.S. banks Canadian banks
4.4 The 1997 structural break and the development of the feedback effect
Figure 10 shows that a structural break occurred in the banks’ non-interest income series around
1997, both in the U.S. and in Canada. The Quandt-Andrews unknown breakpoint test10 reveals that a
breakpoint is depicted for the U.S. snonin (share of non-interest income in total net operating income)
series around the third quarter of 1997, and around the first quarter of 1997 for the Canadian banks’
corresponding time series11. In this section, we study the incidence of this structural break on our VAR
results. Accordingly, we thus recast our analysis over two subperiods: 1984-1996 and 1997-2013.
10 For more detail on this test, see Quandt (1960), Andrews (1993, 2003), and Stock and Watson (2003, 2011). According to Stock and Watson (2011, p. 560), the QLR (Quandt likelihood ratio) statistic is given by:
( ) ( ) ( )0 0 1 0 1, 1 ,...,QLR MAX F F Fτ τ τ τ τ τ= + ≤ ≤⎡ ⎤⎣ ⎦ where F(.) refers to the standard F statistic evaluated at time
τ . In
other words, the QLR statistic is the maximum F statistic computed over a possible set of breakpoints stretched over the sample. It is thus a generalization of the basic Chow test. 11 For more detail on this structural break, see Calmès and Théoret (2010, 2014). This break is associated with a consolidation of the growth of non-interest income in Canada and in the U.S. It is also related to the adoption of the VaR by banks as a gauge of market risk. Finally, a risk premium was added to Canadian bank returns around 1997 which account for the greater risk embedded in bank non-traditional business lines (Calmès and Théoret, 2010).
0
50
100
150
200
250
300
350
82 84 86 88 90 92 94 96 98 00 02 04 06 08 10 12
QLR=338
1% critical value = 164
breakpoint: 1997q3
f sta
tistic
0
100
200
300
400
500
82 84 86 88 90 92 94 96 98 00 02 04 06 08 10 12
breakpoint: 1997q1 QLR=489
1% critical value = 240
22
Figure 11 Feedback effect from banks’ fee-based and traditional activities to real GDP on two subperiods: 1984-1996 and 1997-2013.
Panel A: U.S.
1984-1996
1997-2013
Panel B: Canada
1984-1996
-1.0
-0.5
0.0
0.5
1.0
1.5
2 4 6 8 10 12 14 16 18 20
response of real gdp growth to non-interest income growth
-1.0
-0.5
0.0
0.5
1.0
1.5
2 4 6 8 10 12 14 16 18 20
response of real GDP growth to net interest income growth
-1.0
-0.5
0.0
0.5
1.0
1.5
2 4 6 8 10 12 14 16 18 20
response of GDP growth to non-interest income growth
-1.0
-0.5
0.0
0.5
1.0
1.5
2 4 6 8 10 12 14 16 18 20
response of GDP growth to net interest income growth
-2
-1
0
1
2
3
2 4 6 8 10 12 14 16 18 20
response of real GDP growth to non-interest income growth
-2
-1
0
1
2
3
2 4 6 8 10 12 14 16 18 20
response of real GDP growth t net interest income growth
23
1997-2013
Before 1997, there is no significant feedback effect from U.S. bank income growth to GDP
growth (Figure 11, Panel A). The situation materializes only during the last period (1997-2013): a
significant positive feedback effect from non-interest income growth to GDP growth appears after 1997.
A significant negative feedback effect is also observed for net interest income growth, although it is much
weaker.
In Canada, there is a significant positive feedback effect from non-interest income growth to GDP
growth over the period 1984-1996, but it clearly gains strength over the second period (1997-2013)
(Figure 11, Panel B). Recall that universal banking has developed since 1987 in Canada, which might
explain the earlier presence of the feedback effect. On the other hand, there is no feedback effect from net
interest income growth to GDP growth from 1984 to 1996. However, similarly to the U.S., a significant
negative feedback effect has emerged since 1997.
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2 4 6 8 10 12 14 16 18 20
response of real GDP growth to non-interest income growth
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2 4 6 8 10 12 14 16 18 20
response of real GDP growth to net interest income growth
24
Figure 12 Feedback effect from banks’ fee-based and traditional activities to the stock market on two subperiods: 1984-1996 and 1997-2013.
Panel A: U.S.
1984-1996
1997-2013
-10
-5
0
5
10
15
2 4 6 8 10 12 14 16 18 20
response of the stock market return to non-interest income growth
-10
-5
0
5
10
15
2 4 6 8 10 12 14 16 18 20
response of the stock market return to net interest income growth
-15
-10
-5
0
5
10
15
2 4 6 8 10 12 14 16 18 20
response of the stock market return to non-interest income growth
-15
-10
-5
0
5
10
15
2 4 6 8 10 12 14 16 18 20
response of the stock market return to net interest income growth
25
Panel B: Canada
1984-1996
1997-2013
-10
-5
0
5
10
15
2 4 6 8 10 12 14 16 18 20
response of the stock market return to non-interest income growth
-10
-5
0
5
10
15
2 4 6 8 10 12 14 16 18 20
response of the stock market return to net interest income growth
-10
-5
0
5
10
15
2 4 6 8 10 12 14 16 18 20
response of the stock market return to non-interest income growth
-10
-5
0
5
10
15
2 4 6 8 10 12 14 16 18 20
response of the stock market return to net interest income growth
26
The feedback effect from bank income growth to the stock market has a similar pattern in both
countries (Figure 12, Panels A and B). In the U.S., a significant positive feedback effect from bank non-
interest income growth to the stock market appears over the last period. This feedback effect has
developed earlier in Canada, but it is also stronger over the second period.
Figure 13 Response of net interest income growth to the lagged values of macroeconomic and financial shocks on two subperiods: 1984-1996 and 1997-2013
Panel A: U.S. GDP growth stock market return short-term interest rate
Panel B: Canada GDP growth stock market return short-term interest rate
-.3
-.2
-.1
.0
.1
.2
.3
.4
.5
.6
5 10 15 20 25 30 35
1984-1996 1997-2013
Cross-correlation: net interest income growth and GDP growth
-.3
-.2
-.1
.0
.1
.2
.3
.4
.5
5 10 15 20 25 30 35
1984-1996 1997-2013
Cross-correlation: net interest income growth and stock market return
-.5
-.4
-.3
-.2
-.1
.0
.1
.2
.3
.4
5 10 15 20 25 30 35
1984-1996 1997-2013
Cross-correlation: net interest income growth and short-term interest rate
-.6
-.4
-.2
.0
.2
.4
5 10 15 20 25 30 35
1984-1996 1997-2013
Cross-correlation: net income growth and GDP growth
-.8
-.6
-.4
-.2
.0
.2
.4
5 10 15 20 25 30 35
1984-1996 1997-2014
Cross-correlation: net interest income growth and stock market return
-.6
-.5
-.4
-.3
-.2
-.1
.0
.1
.2
.3
5 10 15 20 25 30 35
1984-1996 1997-2013
Cross-correlation: net interest income growth and short-term interest rate
27
4.5 The behaviour of net interest income growth before and after the structural break
Since monetary policy essentially impacts the banking system through net interest income, it is
interesting to examine how the behaviour of net interest income growth might have changed before
versus after the structural break. Based on the cross-correlations analysis, U.S. banks’ net interest income
growth is much more responsive to GDP growth and to the stock market after the 1997 structural break
(Figure 13, Panel A)12. Moreover, the sensitivity of U.S. banks’ net interest income growth to the short-
term interest rate does not decrease after the structural break. This result might seem counterintuitive
since more financial instruments working as insurance vehicles (like credit derivatives) are available
during the last period. The fact that banks remain evenly exposed to monetary shocks despite these
vehicles is broadly consistent with the argument that financial innovation is mainly used to take more
calculated risks rather than to protect the banking business (Demsetz and Strahan, 1997).
In Canada, the sensitivity of net interest income growth to GDP growth and the stock market
turns from positive to negative in the short-run after the structural break (Figure 13, Panel B). This may
be due to the negative correlation between Canadian bank net interest and non-interest income flows, the
latter being in stark expansion during the second period. As in the U.S., net interest income growth also
reacts more negatively to a positive interest rate shock in the short-run after the structural break13.
Figure 14 IRFs: response of net interest income growth to various shocks over two subperiods: 1984-1996 and 1997-2014
Panel A: U.S. 1984-1996
GDP growth stock market return short-term interest rate
1997-2013 GDP growth stock market return short-term interest rate
12 This may be attributable to increasing complementarities between bank traditional and non-traditional activities. 13 It recuperates very quickly thereafter, a pattern which may be explained by the higher weight of wholesale funds in Canadian banks’ funding after the structural break.
-4
-3
-2
-1
0
1
2
3
4
2 4 6 8 10 12 14 16 18 20
Response of net interest income growth to GDP growth1984-1996
-2
-1
0
1
2
3
2 4 6 8 10 12 14 16 18 20
Response of net interest income growth to stock market return1984-1996
-3
-2
-1
0
1
2
3
2 4 6 8 10 12 14 16 18 20
response of net interest income growth to short-term interest rate1984-1996
-2
-1
0
1
2
3
2 4 6 8 10 12 14 16 18 20
Response of net interest income growth to real GDP growth1997-2013
-2
-1
0
1
2
3
2 4 6 8 10 12 14 16 18 20
response of net interest income growth to stock market return1997-2013
-2
-1
0
1
2
3
2 4 6 8 10 12 14 16 18 20
Response of net interest income growth to short-term interest rate1997-2013
28
Panel B: Canada 1984-1996
GDP growth stock market return short-term interest rate
1997-2013 GDP growth stock market return short-term interest rate
The IRFs of net interest income growth support our preliminary results. In the U.S., the short-
term response of net interest income growth to GDP growth and to the stock market turns from positive
to negative after the structural break (Figure 14, Panel A). More importantly, the negative short-term
impact of a monetary shock on net interest income growth shows no obvious change.
In Canada, the response of net interest income growth to GDP growth and the stock market has
switched from positive to negative even more (Figure 14, Panel B), and again, the IRFs do not signal any
change in the negative relationship between net interest income growth and interest rate shocks.
To summarize, bank net interest income growth has become countercyclical, at least in the short-
run, both in Canada and in the U.S. In other words, net interest income seems to act as a buffer against
fluctuations since the 1997 structural break14. A positive shock originating from the stock market also
tends to decrease net interest income growth in both countries. That may be explained by the “search-for-
yield” effect on the side of bank depositors, which tends to increase banks’ cost of funds and decrease net
interest income. Finally, monetary policy does not seem to have lost its effectiveness over the behaviour
of net interest income.
5. Conclusion
Prior to the advent of universal banking, banks were based on the originate-to-hold model, and
loan growth was strongly linked to GDP growth. However more recently, counteracting forces seem to
have surfaced. On the one hand, fee-based activities offer greater diversification opportunities. With the
14 As Calmès and Théoret (2014) suggest however, this might be due to a simple risk shifting, as the detrimental impact of OBS activities seem to have increased, whilst credit risk was becoming more manageable.
-4
-3
-2
-1
0
1
2
3
4
2 4 6 8 10 12 14 16 18 20
response of net interest income growth to real GDP growth1984-1996
-4
-2
0
2
4
6
2 4 6 8 10 12 14 16 18 20
response of net interest income growth to stock market return1984-1996
-4
-3
-2
-1
0
1
2
3
4
2 4 6 8 10 12 14 16 18 20
Response of net interest income growth to short-term interest rate1984-1996
-6
-4
-2
0
2
4
6
2 4 6 8 10 12 14 16 18 20
Response of net interest income growth to real GDP growth1997-2013
-6
-4
-2
0
2
4
6
2 4 6 8 10 12 14 16 18 20
Response of net interest income growth to stock market return1997-2013
-6
-4
-2
0
2
4
6
2 4 6 8 10 12 14 16 18 20
Response of net interest income growth to short-term interest rate1997-2013
29
accelerated development of derivatives—especially credit derivatives like credit swaps—banks can better
manage credit and market fluctuations. On the other hand, diversification in fee-based activities increases
the volatility of banks’ income flows and, ceteris paribus, this factor tends to increase the amplitude of the
business cycle. The main contribution of this paper is to show that such a feedback effect is indeed at play
in the data.
In the U.S., given banks’ product-mix, lagged values of GDP growth still have a significant
impact on non-interest income growth. However, non-interest income growth seems to give rise to
significant feedback effects, both on real GDP and on the stock market, a result much in line with
Peersman and Wagner (2014). In particular, we show that these feedback effects gained strength after the
1997 structural break in banks’ share of non-interest income in net operating income.
In the universal banking era, banks might have a greater impact on economic activity and on the
stock market but central banks nevertheless continue to impact net interest income growth, although this
effect seems to have become shorter-lived.
30
References
Andrews, D., 2003. Tests for parameter instability and structural change with unknown change
point: A corrigendum. Econometrica 71, 395-397. Bernanke, B., Gertler, M., Gilchrist, S., 1999. The financial accelerator in a quantitative business
cycle framework. In: Taylor, J., Woodford, M. (eds.). Handbook of Macroeconomics. Bikker, J., Hu, H., 2002. Cyclical patterns in profits provisioning and lending of banks and pro-
cyclicality of the new Basel capital requirements. Working Paper, Netherlands Central Bank.
Brown, G., 2010. Beyond the Crash: Overcoming the First Crisis of Globalization. Free Press, New York.
Calmès, C., Liu, Y., 2009. Financial structure change and banking income: A Canada – U.S. comparison. Journal of International Financial Markets, Institutions and Money 19, 128-139.
Calmès, C., Théoret, R., 2010. The impact of off-balance-sheet activities on banks returns: An application of the ARCH-M to Canadian data. Journal of Banking and Finance 34, 1719-1728.
Calmès, C., Théoret, R., 2013. Market-oriented banking, financial stability and macro-prudential indicators of leverage. Journal of International Financial Markets, Institutions and Money, 27, 13-34.
Calmès, C., Théoret, R., 2014. Bank systemic risk and macroeconomic shocks: Canadian and U.S. evidence. Journal of Banking and Finance 40, 388-402.
Calmès, C., Théoret, R., 2015. Product-mix and bank performance: New U.S. and Canadian evidence. Managerial Finance, forthcoming.
DeJong, D.N., Dave, C., 2007. Structural macroeconometrics. Princeton University press. Princeton.
Demsetz, R.S., Strahan, P.E., 1997. Diversification, size, and risk at bank holding companies. Journal of Money, Credit and Banking 29, 300–313.
Dewatcher, H.D., Wouters, R., 2014. Endogenous risk in a DSGE model with capital constrained financial intermediaries. Journal of Economic Dynamics and Control 43, 241-268.
Gennaioli, N., Shleifer, A., Vishny, R.W., 2013. A model of shadow banking. Journal of Finance 68, 1331-1363.
Gertler, M., Kiyotaki, N., 2011. Financial intermediation and credit policy in business cycle analysis. In: Friedman, B., Woodford, M. (eds.). Handbook of Monetary Economics. Elsevier, Amsterdam.
Hamilton, J.D., 1994. Time Series Analysis. Princeton University Press, Princeton (New Jersey). Jacobson, T., Lindé, J., Roszback, K., 2005. Exploring interactions between real activity and the
financial crisis. Journal of Financial Stability 1, 308-341. Kiyotaki, N., Moore, J., 1997. Credit cycles. Journal of Political Economy 105, 211-248. Marcucci, J., Quagliariello, M. 2006. Is bank portfolio riskiness procyclical? Evidence from Italy
using a vector autoregression. Journal of International Financial Markets, Institutions and Money 18, 46-63.
Peersman, G., Wagner, W., 2014. Shocks to bank lending, risk-taking, securitization, and their role for U.S. business fluctuations. Working Paper, European Banking Center.
Pesaran, M.H., Shin, Y., 1998. Generalized impulse response analysis in linear multivariate models. Economics Letters 58, 17-29.
Quandt, R., 1960. Tests of the hypothesis that a linear regression system obeys two separate regimes. Journal of the American Statistical Association 55, 324-330.
Saunders, A., Cornett, M.M., McGraw, P.A., 2014. Financial Institutions Management: A Risk Management Approach. Fifth Canadian Edition. McGraw Hill.
Shleifer, A., Vishny, R., 2010. Asset fire sales and credit easing. American Economi Review: papers and Proceedings, 46-50.
Sims, C., 1980. Macroeconomics and reality. Econometrica 48, 1-48. Stiroh, K.J., 2004. Diversification in banking: Is noninterest income the answer? Journal of Money,
Credit and Banking 36, 853-882.
31
Stock, J.H., Watson, M.W., 2003. Has the business cycle changed? And why? In: NBER, Macroeconomics Annual 2002, vol. 17, MIT Press.
Stock, J.H., Watson, M.W., 2011. Introduction to Econometrics. 3rd edition. Addison-Wesley Pearson.
Williamson, S.D., 1987. Financial intermediation, business failures and real business cycles. Journal of Political Economy 95, 1196-1216.
32